They are likely stuck in old metadata, where the bucket needs to be optimized. I suggest you first try to alter your search to workaround the problem. Secondly, when your deleted data gets frozen/deleted, this problem will go away.
The reason why you are probably seeing this, is that the metadata still exists for that host. It's possible it is a bug, but there are ways to check why this is occurring. You may want to run the following search to see if it is a metadata problem:
| metadata type=hosts host=<your_host_with_zero_events>
If your host shows up in the results with a totalCount that is incorrect, then there might be a problem. You should probably log a support case at that point.
... View more